The Badge Economy Is Not Building AI Capability

Eighty-seven per cent of digital workers now use AI at work.

Seventy-five per cent say it makes them more productive.

Only 13 per cent say it has significantly improved company performance.

That is not just a statistic, but what is fundamentally wrong in the way we are enabling Employees on AI.

If you are a CEO, CHRO, CMO, or CIO sitting with those numbers from the Work AI Index 2026 by Glean’s Work AI Institute, you should be asking one uncomfortable question: what exactly is our current AI training setup, vendors, modules, how specific is it and is it designed for our organisation?

Because what I am seeing across organisations in APAC is not an AI capability problem. It is an execution-fluency problem stemming from a poor vision of how best to train and empower employees.

Course Completion Is Not Capability

Just for clarity, and to make sure my point comes across. I am not against training. AI literacy matters is probably the #1 priority for any organisation. Responsible experimentation matters. Helping your people understand what these tools can and cannot do is genuinely important, especially in a region where regulatory diversity, language fragmentation, and governance maturity vary so sharply across markets.

But here is what most organisations are doing instead of building capability: they are ticking boxes, creating employee leaderboards and manufacturing credentials.

Badges. Bootcamps. Micro-credentials. Prompt engineering courses. AI awareness workshops. All of it tracked, all of it announced, none of it connected to whether the work actually got better.

A badge does not tell you whether your sales director can use AI to sharpen account prioritisation.

A certificate does not tell you whether your marketing leader can use AI to improve segmentation quality or reduce the time between insight and action.

A workshop attendance report tells you nothing about whether a finance manager can safely interrogate cost drivers using AI, let alone explain the output to a sceptical CFO.

What we have done is confuse learning activity with the acquisition of relevant capability, and motion with momentum.

Most level-headed consulting reports stress that simply inserting AI tools into existing ways of working is not enough. The value comes when businesses reshape workflows end-to-end. Not the tools. The workflows.

That is the part most organisations avoid because workflow redesign is harder than buying licences. It requires governance. It requires manager capability. It requires people to make choices and be accountable for them. Training is easy to announce. Behaviour change is hard to deliver. And so we keep announcing new training modules.

The People Feeling This Most Are the People You Cannot Afford to lose.

There is a specific group absorbing most of the pressure here, and they are not the ones getting the most attention.

Not the junior employees being positioned as AI natives.

Not the C-suite making the transformation announcements.

It is the mid-career professionals and upper-middle managers sitting in the operational middle. They carry the execution burden. They manage the teams. They translate strategy into the weekly work that actually serves customers. They absorb pressure from above and anxiety from below, and now they are being told to “use AI” while simultaneously protecting quality, productivity, compliance, team morale, customer trust, and their own professional relevance.

That is a significant ask.

In many APAC organisations, the tension is sharper still. A regional leader might be expected to deploy AI across Singapore, India, Indonesia, Australia, Japan, and parts of Southeast Asia simultaneously, each with different data regulations, languages, customer behaviours, and governance readiness. A generic AI certification solves none of that. It creates awareness. In no way does it come close to creating workflow execution fluency.

And here is the cruel irony: the people most capable of driving real workflow change are the ones who have the least time and operational headroom to experiment. So they complete the course, return to the same workflow, and quietly conclude that AI is just one more thing on their plate.

That is how transformation becomes gimmicky. It’s the classic digital transformation failure story.

The Hidden Labour Nobody Is Measuring

There is something else worth naming, because I rarely see it come up in the AI adoption conversation.

AI does not always remove work. Sometimes it changes the shape of work.

Glean’s research calls this “botsitting.” The hidden labour of making AI usable: feeding context into tools, checking outputs that sound confident but are wrong, debugging prompts, rerunning queries, cleaning up hallucinations before they reach a client or a board deck. This work sits with the people who are already overloaded.

So before your organisation celebrates high AI usage numbers, ask the harder question: is usage translating to value, or are your best people now managing a machine on top of everything else they were already managing?

High usage is not a success metric. Measurably better work is.

The Real Deficit Is Judgment Infrastructure

In thirty years of working and building regional teams, Centres of Excellence, and revenue-linked programmes across Asia Pacific, one lesson has stayed constant regardless of the technology cycle.

Clarity is a leadership responsibility.

People do not become capable because we tell them the future is changing. They become capable when they are given a clear operating context, safe boundaries, real problems to solve, and permission to practise. Remove any one of those four elements, and you get either paralysis or shadow adoption, people using AI tools off the books because the approved path is too slow or too vague.

Neither outcome serves you.

The shift that organisations need to make is not from “no training” to “more training.” It is from generic AI literacy to workflow-specific fluency. From “everyone must learn prompting” to “this team must improve this specific business process using AI, responsibly, by this date.”

That is a very different programme. And it starts with leadership making deliberate choices, not just setting ambitious transformation targets.

What Actually Works: Five Moves, From the Room Where It Happened

Here is what I would do instead of another enterprise-wide AI training rollout. These are not hypotheses. They are patterns I have seen work, and the conditions under which they work.

Start with friction, not tools.

I have sat in enough strategy sessions where the agenda item is “AI tool rollout” before anyone has answered the question of what problem we are solving. The sequence is backwards. The tool is not the starting point. The pain is.

Ask the team where time is haemorrhaging. Where is the quality inconsistent week to week? Where does a decision that should take two days take two weeks because the right insight is buried in a format nobody has time to read? Where is the handoff between functions breaking down so reliably that everyone has accepted it as normal?

That is where AI belongs. Not as a curiosity. As a targeted intervention in a real operational problem that someone is already losing sleep over.

When you start there, the training question answers itself. You are not asking people to “learn AI.” You are asking them to fix something they already want fixed.

Give people a micro-assignment, not a mandate.

In my experience, when we were building out the first ABM framework for Asia Pacific, we did not hand the region a playbook and tell them to execute. We gave small teams a specific account, a specific problem, and a specific deadline. The constraint was what made it real.

The same logic applies here. Do not send your marketing director to a two-day AI bootcamp and expect transformation on Monday. Give them one assignment: use AI to improve the quality of your top ten account prioritisation decisions before the next QBR. That is small enough to complete in a week. Important enough that the result will be noticed. Specific enough to evaluate the output, not just the effort.

The assignment has to matter commercially. If it does not matter, people will treat it like every other corporate initiative that was urgent last quarter and forgotten this quarter.

Put the guardrails up before anyone starts driving.

This is where I see organisations make the most costly mistake. They deploy the tools, run the training, and leave governance as a future agenda item. Then something goes wrong: an AI-generated output reaches a client without human review, sensitive data gets fed into an unapproved tool, a confident-sounding but factually wrong recommendation lands in a board deck, and the response is to restrict everything. Enthusiasm becomes compliance theatre in reverse.

I built the Guardians of Trust framework precisely because I kept watching this cycle repeat. Governance is not a brake on AI adoption. It is the architecture that enables sustainable adoption. Before any micro-assignment begins, the team needs to know four things clearly: what data is in scope, which tools are approved, what outputs require a human checkpoint before they move forward, and who to call when something feels wrong. That is not a bureaucratic checklist. That is the trust infrastructure that allows people to experiment without fear and to know that the organisation has their back when edge cases arise.

In APAC especially, this cannot be a one-size document. The data governance environment in Singapore is not the same as in Indonesia. The cultural comfort with flagging an error differs between markets. Get specific, or the framework becomes noise.

Measure the work, not the hours.

I once had a potential client who was genuinely proud that their team had logged over a certain number of hours of AI training in a single quarter. When I asked what had changed in their workflows, the conversation abruptly ended.

Many training hours logged in. Nothing measurably different.

The question that should be on every leader’s dashboard is not how many people completed the training. It is: which workflows became demonstrably better after AI was introduced? Did the time between customer insight and campaign response shrink? Did the quality of account scoring improve enough for sales to start trusting the marketing input again? Did the finance team catch a budget variance pattern two weeks earlier than they otherwise would have?

If you cannot point to a specific process that is faster, sharper, or less error-prone, you do not have AI adoption. You have AI attendance.

Coach the judgment, not just the tool.

This is the one most organisations skip, and it is the one that determines whether the other four stick.

A mid-career leader who completes an AI course knows how to use a prompt. That is not the hard part. The hard part is judgment. Knowing when to use AI and when not to. Recognising the difference between an output that is good enough to act on and one that sounds plausible but needs another pass. Being able to walk a sceptical CFO or a cautious regional MD through an AI-assisted recommendation without losing the room.

These are not tool skills. They are leadership skills applied in a new context. And they develop through practice, feedback, and honest conversation with someone who has been in those rooms.

If you are investing in AI tools and AI training but not in the coaching layer that builds judgment, you are building a capability with no steering wheel. The people you most need to retain are the ones who will feel this gap first and disengage fastest when nobody addresses it.

The Real Risk Is Not Low Completion Rates

The real risk is not that your employees fail to complete their AI training modules.

The real risk is that they complete everything, feel no clearer, return to the same workflow, and quietly decide that AI is the next version of the enterprise software rollout that promised transformation and delivered confusion.

That is how fatigue builds. That is how trust erodes. That is how your best execution talent, the people who are actually capable of driving change, starts to disengage.

And if you are wondering what that costs, it costs more than the AI licensing budget. It costs the human capital that cannot be re-licensed next year.

For the Builders and Decision Makers Reading This

Your people want to grow. Most of them are not resisting AI out of fear of technology. They are resisting confusion. They are resisting vague mandates. They are resisting the exhaustion of being asked to prepare indefinitely without being allowed to perform differently.

So before you approve the next enterprise AI training budget, sit with a harder question first.

Have we built the internal architecture, governance, workflow permissions, manager coaching, and measurement systems so people can actually use what they learn?

If the answer is not a confident yes, the next badge will not solve the problem.

It will simply add one more credential to the burnout pile.